Artificial neural networks (ANN), adaptive neuro-fuzzy inference system (ANFIS) with subtractive clustering (ANFIS-SC) and support vector machine models were used to determine total dissolved solids (TDS) of the Zayandehrood River in Iran. In total, nine hydrochemical parameters [Ca2+, SO42−, Na+, Cl−, EC, pH, HCO3−, Mg2+ and sodium adsorption ratio (SAR)] were utilized to estimate the TDS of the river at a monthly time scale. Statistical data were categorized into low-flow and wet periods based on river discharge. Principal component analysis (PCA) was used to determine the input of the models. The results indicate that the PCA method, in both wet and low-flow periods, performed suitably based on the evaluation criteria for all models. The parameters of the first component included Ca2+, SO42−, Cl−, EC, Mg2+ and SAR in both periods. In contrast, the parameters pH and HCO3− of the second component provided unacceptable precision. The ANFIS-SC model was more precise than the other two models, with an RMSE value of 12.33 meq/l for the first component in the low-flow period. However, the ANN model was most precise in the wet period, with a calculated RMSE value of 13.87 meq/l.
|Journal||Sustainable Water Resources Management|
|State||Published - Apr 2021|
Bibliographical notePublisher Copyright:
© 2021, The Author(s), under exclusive licence to Springer Nature Switzerland AG part of Springer Nature.
- Qualitative parameters
ASJC Scopus subject areas
- Renewable Energy, Sustainability and the Environment
- Water Science and Technology